New TU Delft Drone Uses Touch for Quieter, Longer Flights
A new drone developed by researchers at TU Delft in the Netherlands employs a tactile sensing system, allowing it to perch on branches using touch rather than vision. This approach seeks to enable quieter operation and extended battery life, addressing limitations of conventional drones. The device detects branch contact, ascertains orientation, adjusts its position, and grips it securely, mimicking avian landing behaviours.
Traditional drones often produce noise during hover, limiting their use. Their power cells deplete quickly with continuous flight. Operating in dense environments, such as rainforest canopies, presents difficulties for drones reliant on vision, as grippers and foliage obscure camera views. This new TU Delft development aims to circumvent these prevalent issues.
But this novel drone functions more akin to a bird, identifying a branch through physical contact before settling. Embedded with soft tactile sensors within an anthropomorphic hand, the drone senses when it encounters a branch, discerning its precise location and angle. It then repositions itself, secures its grip, and deactivates its motors, conserving energy. According to the research published in *npj Robotics*, this represents a new method for autonomous perching, where touch becomes the primary source of operational information.
Existing drone systems have typically depended on visual data to interpret their surroundings, requiring an unobstructed perspective or a detailed model of a branch. Such reliance renders these drones effectively blind during the critical final moments of approach, as their grippers or the target branch itself obstruct the camera's field of view. Most studies focusing on vision based perching prove impractical for real world deployment.
This tactile drone refines an approximate target location through physical interaction. It executes a figure eight search, opening and closing its hand with three fingers. Initial contact provides branch position and orientation data. The drone rotates and adjusts its stance until three finger sensors confirm stable contact, switching off its motors. Failure to grasp prompts the drone to hover and reattempt.
And natural animal behaviours, particularly how birds use touch to guide landings and adjust grip, inspired this drone's conceptual basis. Associate Professor Dr Salua Hamaza, TU Delft, aimed to integrate this capability into aerial robots. She explained visual systems inform about object presence until engagement, when the gripper impedes view. Animals bypass this by using touch, closing the loop. This was the drone's objective: to enter unseen spaces, ascertain contents through touch, and adjust while in contact.
A light, anthropomorphic hand engineered for flight forms the outcome. It has three fingers, each with three phalanges, inspired by human hand proportions. Torsional springs close fingers naturally, maintaining grip without power once perched. A single tendon opens each finger; soft silicone pads provide friction and adapt to surfaces.
But each phalange incorporates a tactile sensor, a small copper electrode beneath the silicone surface. Contact with an object changes the electrical signal, producing a binary contact signal. Individually, this signal offers limited information. The drone's knowledge of its hand's form and finger positions allows it to accurately determine each touch location. This simple touch signal translates into a spatial point and a movement direction.
So Anton Bredenbeck, TU Delft, stated each sensor merely indicates presence. Understanding the hand's geometry, he explained, allows reconstruction of the branch's orientation and position. The drone builds an accurate target picture through physical interaction.
This method exhibits robustness by using touch to rectify perceptual errors. Flight tests showed consistent reliability even with incorrect estimations for target position, orientation, and size. A standard vision reliant approach failed rapidly in similar conditions. This research suggests drones capable of sustained environmental monitoring, especially where visual data is insufficient, such as forest canopies, industrial frameworks, or areas without motion capture systems.
The study advocates re-evaluating contact in aerial robotics. Instead of an inherent hazard, it proposes contact as a valuable data source to guide actions. Future drones, like their natural counterparts, may rely on tactile feedback to navigate complex environments and identify unseen landing sites.
TU Delft developed a drone utilising tactile sensors for perching.
This drone uses a hand with three fingers to detect and grip branches.
Tactile perching offers advantages in quiet operation and extended battery life over vision based systems.
The method demonstrates robustness in complex environments, even with inaccurate target estimations.
The approach redefines physical contact in aerial robotics as an information source, not a liability. Source: techxplore
